path: root/erts/doc/src/erts_alloc.xml

<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE cref SYSTEM "cref.dtd">

<cref>
  <header>
    <copyright>
      <year>2002</year><year>2018</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      Licensed under the Apache License, Version 2.0 (the "License");
      you may not use this file except in compliance with the License.
      You may obtain a copy of the License at
 
          http://www.apache.org/licenses/LICENSE-2.0

      Unless required by applicable law or agreed to in writing, software
      distributed under the License is distributed on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
      See the License for the specific language governing permissions and
      limitations under the License.

    </legalnotice>

    <title>erts_alloc</title>
    <prepared>Rickard Green</prepared>
    <docno>1</docno>
    <date>2003-06-11</date>
    <rev>1</rev>
    <file>erts_alloc.xml</file>
  </header>
  <lib>erts_alloc</lib>
  <libsummary>An Erlang runtime system internal memory allocator library.
  </libsummary>
  <description>
    <p><c>erts_alloc</c> is an Erlang runtime system internal memory
      allocator library. <c>erts_alloc</c> provides the Erlang
      runtime system with a number of memory allocators.</p>
  </description>

  <section>
    <title>Allocators</title>
    <marker id="allocators"></marker>
    <p>The following allocators are present:</p>

    <taglist>
      <tag><c>temp_alloc</c></tag>
      <item>Allocator used for temporary allocations.</item>
      <tag><c>eheap_alloc</c></tag>
      <item>Allocator used for Erlang heap data, such as Erlang process heaps.
      </item>
      <tag><c>binary_alloc</c></tag>
      <item>Allocator used for Erlang binary data.</item>
      <tag><c>ets_alloc</c></tag>
      <item>Allocator used for <c>ets</c> data.</item>
      <tag><c>driver_alloc</c></tag>
      <item>Allocator used for driver data.</item>
      <tag><c>literal_alloc</c></tag>
      <item>Allocator used for constant terms in Erlang code.</item>
      <tag><c>sl_alloc</c></tag>
      <item>Allocator used for memory blocks that are expected to be
        short-lived.</item>
      <tag><c>ll_alloc</c></tag>
      <item>Allocator used for memory blocks that are expected to be
        long-lived, for example, Erlang code.</item>
      <tag><c>fix_alloc</c></tag>
      <item>A fast allocator used for some frequently used
       fixed size data types.</item>
      <tag><c>std_alloc</c></tag>
      <item>Allocator used for most memory blocks not allocated through any of
        the other allocators described above.</item>
      <tag><c>sys_alloc</c></tag>
      <item>This is normally the default <c>malloc</c> implementation
        used on the specific OS.</item>
      <tag><c>mseg_alloc</c></tag>
      <item>A memory segment allocator. It is used by other
        allocators for allocating memory segments and is only
        available on systems that have the <c>mmap</c> system
        call. Memory segments that are deallocated are kept for a
        while in a segment cache before they are destroyed. When
        segments are allocated, cached segments are used if possible
        instead of creating new segments. This to reduce
        the number of system calls made.</item>
    </taglist>

    <p><c>sys_alloc</c>, <c>literal_alloc</c> and <c>temp_alloc</c> are always
      enabled and cannot be disabled. <c>mseg_alloc</c> is always enabled if it
      is available and an allocator that uses it is enabled. All other
      allocators can be <seecref marker="#M_e">enabled or disabled</seecref>.
      By default all allocators are enabled.
      When an allocator is disabled, <c>sys_alloc</c> is used instead of
      the disabled allocator.</p>

    <p>The main idea with the <c>erts_alloc</c> library is to separate
      memory blocks that are used differently into different memory
      areas, to achieve less memory fragmentation. By
      putting less effort in finding a good fit for memory blocks that
      are frequently allocated than for those less frequently
      allocated, a performance gain can be achieved.</p>
  </section>

  <section>
    <marker id="alloc_util"></marker>
    <title>The alloc_util Framework</title>
    <p>Internally a framework called <c>alloc_util</c> is used for
      implementing allocators. <c>sys_alloc</c> and
      <c>mseg_alloc</c> do not use this framework, so the
      following does <em>not</em> apply to them.</p>

    <p>An allocator manages multiple areas, called carriers, in which
      memory blocks are placed. A carrier is either placed in a
      separate memory segment (allocated through <c>mseg_alloc</c>), or in
      the heap segment (allocated through <c>sys_alloc</c>).</p>

    <list type="bulleted">
      <item>
        <p>Multiblock carriers are used for storage of several blocks.</p>
      </item>
      <item>
        <p>Singleblock carriers are used for storage of one block.</p>
      </item>
      <item>
        <p>Blocks that are larger than the value of the singleblock carrier
          threshold (<seecref marker="#M_sbct"><c>sbct</c></seecref>) parameter
          are placed in singleblock carriers.</p>
      </item>
      <item>
        <p>Blocks that are smaller than the value of parameter <c>sbct</c>
          are placed in multiblock carriers.</p></item>
    </list>

    <p>Normally an allocator creates a "main multiblock
      carrier". Main multiblock carriers are never deallocated. The
      size of the main multiblock carrier is determined by the value of
      parameter <seecref marker="#M_mmbcs"><c>mmbcs</c></seecref>.</p>

    <p><marker id="mseg_mbc_sizes"></marker>Sizes of multiblock carriers
      allocated through <c>mseg_alloc</c> are decided based on the
      following parameters:</p>

    <list type="bulleted">
      <item>The values of the largest multiblock carrier size
        (<seecref marker="#M_lmbcs"><c>lmbcs</c></seecref>)</item>
      <item>The smallest multiblock carrier size
        (<seecref marker="#M_smbcs"><c>smbcs</c></seecref>)</item>
      <item>The multiblock carrier growth stages
        (<seecref marker="#M_mbcgs"><c>mbcgs</c></seecref>)</item>
    </list>

    <p>If <c>nc</c> is the current number of multiblock carriers (the main
      multiblock carrier excluded) managed by an allocator, the size
      of the next <c>mseg_alloc</c> multiblock carrier allocated by
      this allocator is roughly
      <c><![CDATA[smbcs+nc*(lmbcs-smbcs)/mbcgs]]></c> when
      <c><![CDATA[nc <= mbcgs]]></c>,
      and <c>lmbcs</c> when <c><![CDATA[nc > mbcgs]]></c>. If the value of
      parameter <c>sbct</c> is larger than the value of parameter
      <c>lmbcs</c>, the allocator may have to create
      multiblock carriers that are larger than the value of
      parameter <c>lmbcs</c>, though.
      Singleblock carriers allocated through <c>mseg_alloc</c> are sized
      to whole pages.</p>

    <p>Sizes of carriers allocated through <c>sys_alloc</c> are
      decided based on the value of the <c>sys_alloc</c> carrier size
      (<seecref marker="#Muycs"><c>ycs</c></seecref>) parameter. The size of
      a carrier is the least number of multiples of the value of
      parameter <c>ycs</c> satisfying the request.</p>

    <p>Coalescing of free blocks are always performed immediately.
      Boundary tags (headers and footers) in free blocks are used,
      which makes the time complexity for coalescing constant.</p>

    <p><marker id="strategy"></marker>The memory allocation strategy
      used for multiblock carriers by an allocator can be
      configured using parameter <seecref marker="#M_as"><c>as</c></seecref>.
      The following strategies are available:</p>

    <taglist>
      <tag>Best fit</tag>
      <item>
        <p>Strategy: Find the smallest block satisfying the
          requested block size.</p>
        <p>Implementation: A balanced binary search tree is
          used. The time complexity is proportional to log N, where
          N is the number of sizes of free blocks.</p>
      </item>
      <tag>Address order best fit</tag>
      <item>
        <p>Strategy: Find the smallest block satisfying the
          requested block size. If multiple blocks are found, choose
          the one with the lowest address.</p>
        <p>Implementation: A balanced binary search tree is
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Address order first fit</tag>
      <item>
        <p>Strategy: Find the block with the lowest address satisfying the
          requested block size.</p>
        <p>Implementation: A balanced binary search tree is
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Address order first fit carrier best fit</tag>
      <item>
        <p>Strategy: Find the <em>carrier</em> with the lowest address that
          can satisfy the requested block size, then find a block within
          that carrier using the "best fit" strategy.</p>
        <p>Implementation: Balanced binary search trees are
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Address order first fit carrier address order best fit</tag>
      <item>
        <p>Strategy: Find the <em>carrier</em> with the lowest address that
          can satisfy the requested block size, then find a block within
          that carrier using the "address order best fit" strategy.</p>
        <p>Implementation: Balanced binary search trees are
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Age order first fit carrier address order first fit</tag>
      <item>
	<p>Strategy: Find the <em>oldest carrier</em> that
          can satisfy the requested block size, then find a block within
          that carrier using the "address order first fit" strategy.</p>
        <p>Implementation: A balanced binary search tree is
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Age order first fit carrier best fit</tag>
      <item>
        <p>Strategy: Find the <em>oldest carrier</em> that
          can satisfy the requested block size, then find a block within
          that carrier using the "best fit" strategy.</p>
        <p>Implementation: Balanced binary search trees are
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Age order first fit carrier address order best fit</tag>
      <item>
        <p>Strategy: Find the <em>oldest carrier</em> that
          can satisfy the requested block size, then find a block within
          that carrier using the "address order best fit" strategy.</p>
        <p>Implementation: Balanced binary search trees are
          used. The time complexity is proportional to log N, where
          N is the number of free blocks.</p>
      </item>
      <tag>Good fit</tag>
      <item>
        <p>Strategy: Try to find the best fit, but settle for the best fit
          found during a limited search.</p>
        <p>Implementation: The implementation uses segregated free
          lists with a maximum block search depth (in each list)
          to find a good fit fast. When the maximum block
          search depth is small (by default 3), this implementation
          has a time complexity that is constant. The maximum block
          search depth can be configured using parameter
          <seecref marker="#M_mbsd"><c>mbsd</c></seecref>.</p>
      </item>
      <tag>A fit</tag>
      <item>
        <p>Strategy: Do not search for a fit, inspect only one free
          block to see if it satisfies the request. This strategy is
          only intended to be used for temporary allocations.</p>
        <p>Implementation: Inspect the first block in a free-list.
          If it satisfies the request, it is used, otherwise a new
          carrier is created. The implementation has a time
          complexity that is constant.</p>
        <p>As from ERTS 5.6.1 the emulator refuses to
          use this strategy on other allocators than <c>temp_alloc</c>.
          This because it only causes problems for other allocators.</p>
      </item>
    </taglist>

    <p>Apart from the ordinary allocators described above, some
      pre-allocators are used for some specific data types. These
      pre-allocators pre-allocate a fixed amount of memory for certain data
      types when the runtime system starts. As long as pre-allocated memory
      is available, it is used. When no pre-allocated memory is
      available, memory is allocated in ordinary allocators. These
      pre-allocators are typically much faster than the ordinary allocators,
      but can only satisfy a limited number of requests.</p>
  </section>

  <section>
    <marker id="flags"></marker>
    <title>System Flags Effecting erts_alloc</title>
    <warning>
      <p>Only use these flags if you are sure what you are
        doing. Unsuitable settings can cause serious performance
        degradation and even a system crash at any time during
        operation.</p>
    </warning>

    <p>Memory allocator system flags have the following syntax:
      <c><![CDATA[+M<S><P> <V>]]></c>,
      where <c><![CDATA[<S>]]></c> is a letter identifying a subsystem,
      <c><![CDATA[<P>]]></c> is a parameter, and <c><![CDATA[<V>]]></c> is the
      value to use. The flags can be passed to the Erlang emulator
      (<seecom marker="erl"><c>erl(1)</c></seecom>) as command-line
      arguments.</p>

    <p>System flags effecting specific allocators have an uppercase
      letter as <c><![CDATA[<S>]]></c>. The following letters are used for
      the allocators:</p>

    <list type="bulleted">
      <item><c>B: binary_alloc</c></item>
      <item><c>D: std_alloc</c></item>
      <item><c>E: ets_alloc</c></item>
      <item><c>F: fix_alloc</c></item>
      <item><c>H: eheap_alloc</c></item>
      <item><c>I: literal_alloc</c></item>
      <item><c>L: ll_alloc</c></item>
      <item><c>M: mseg_alloc</c></item>
      <item><c>R: driver_alloc</c></item>
      <item><c>S: sl_alloc</c></item>
      <item><c>T: temp_alloc</c></item>
      <item><c>Y: sys_alloc</c></item>
    </list>

    <section>
      <title>Flags for Configuration of mseg_alloc</title>
      <taglist>
        <tag><marker id="MMamcbf"/><c><![CDATA[+MMamcbf <size>]]></c></tag>
        <item>
          <p>Absolute maximum cache bad fit (in kilobytes). A segment in the
            memory segment cache is not reused if its size exceeds the
            requested size with more than the value of this
            parameter. Defaults to <c>4096</c>.</p>
        </item>
        <tag><marker id="MMrmcbf"/><c><![CDATA[+MMrmcbf <ratio>]]></c></tag>
        <item>
          <p>Relative maximum cache bad fit (in percent). A segment in the
            memory segment cache is not reused if its size exceeds the
            requested size with more than relative maximum cache bad fit
            percent of the requested size. Defaults to <c>20</c>.</p>
        </item>
        <tag><marker id="MMsco"/><c><![CDATA[+MMsco true|false]]></c></tag>
        <item>
          <p>Sets <seecref marker="#MMscs">super carrier</seecref> only flag.
            Defaults to <c>true</c>. When a super carrier is used and this
            flag is <c>true</c>, <c>mseg_alloc</c> only creates carriers in
            the super carrier. Notice that the <c>alloc_util</c> framework can
            create <c>sys_alloc</c> carriers, so if you want all carriers to
            be created in the super carrier, you therefore want to disable use
            of <c>sys_alloc</c> carriers by also passing
            <seecref marker="#Musac"><c>+Musac false</c></seecref>. When
            the flag is <c>false</c>, <c>mseg_alloc</c> tries to create carriers
            outside of the super carrier when the super carrier is full.</p>
	  <note>
            <p>Setting this flag to <c>false</c> is not supported
              on all systems. The flag is then ignored.</p>
          </note>
        </item>
        <tag><marker id="MMscrfsd"/><c><![CDATA[+MMscrfsd <amount>]]></c></tag>
        <item>
          <p>Sets <seecref marker="#MMscs">super carrier</seecref> reserved
            free segment descriptors. Defaults to <c>65536</c>.
            This parameter determines the amount of memory to reserve for
            free segment descriptors used by the super carrier. If the system
            runs out of reserved memory for free segment descriptors, other
            memory is used. This can however cause fragmentation issues,
            so you want to ensure that this never happens. The maximum amount
            of free segment descriptors used can be retrieved from the
            <c>erts_mmap</c> tuple part of the result from calling
            <seeerl marker="erts:erlang#system_info_allocator_tuple">
            <c>erlang:system_info({allocator, mseg_alloc})</c></seeerl>.</p>
        </item>
        <tag><marker id="MMscrpm"/><c><![CDATA[+MMscrpm true|false]]></c></tag>
        <item>
          <p>Sets <seecref marker="#MMscs">super carrier</seecref> reserve
            physical memory flag. Defaults to <c>true</c>. When this flag is
            <c>true</c>, physical memory is reserved for the whole super
            carrier at once when it is created. The reservation is after that
            left unchanged. When this flag is set to <c>false</c>, only virtual
            address space is reserved for the super carrier upon creation.
            The system attempts to reserve physical memory upon carrier
            creations in the super carrier, and attempt to unreserve physical
            memory upon carrier destructions in the super carrier.</p>
          <note>
            <p>What reservation of physical memory means, highly
              depends on the operating system, and how it is configured. For
              example, different memory overcommit settings on Linux drastically
              change the behavior.</p>
            <p>Setting this flag to <c>false</c> is possibly not supported on
              all systems. The flag is then ignored.</p>
          </note>
        </item>
        <tag><marker id="MMscs"/><c><![CDATA[+MMscs <size in MB>]]></c></tag>
        <item>
          <p>Sets super carrier size (in MB). Defaults to <c>0</c>, that is,
            the super carrier is by default disabled. The super
            carrier is a large continuous area in the virtual address space.
            <c>mseg_alloc</c> always tries to create new carriers in the super
            carrier if it exists. Notice that the <c>alloc_util</c> framework
            can create <c>sys_alloc</c> carriers. For more information, see
            <seecref marker="#MMsco"><c>+MMsco</c></seecref>.</p>
        </item>
        <tag><marker id="MMmcs"/><c><![CDATA[+MMmcs <amount>]]></c></tag>
        <item>
          <p>Maximum cached segments. The maximum number of memory segments
            stored in the memory segment cache. Valid range is <c>[0, 30]</c>.
            Defaults to <c>10</c>.</p>
        </item>
      </taglist>
    </section>

    <section>
      <title>Flags for Configuration of sys_alloc</title>
      <taglist>
        <tag><marker id="MYe"/><c>+MYe true</c></tag>
        <item>
          <p>Enables <c>sys_alloc</c>.</p>
          <note>
            <p><c>sys_alloc</c> cannot be disabled.</p>
          </note>
        </item>
        <tag><marker id="MYtt"/><c><![CDATA[+MYtt <size>]]></c></tag>
        <item>
          <p>Trim threshold size (in kilobytes). This is the maximum amount
            of free memory at the top of the heap (allocated by
            <c>sbrk</c>) that is kept by <c>malloc</c> (not
            released to the operating system). When the amount of free
            memory at the top of the heap exceeds the trim threshold,
            <c>malloc</c> releases it (by calling <c>sbrk</c>).
            Trim threshold is specified in kilobytes.
            Defaults to <c>128</c>.</p>
          <note>
            <p>This flag has effect only when the emulator is linked with
              the GNU C library, and uses its <c>malloc</c> implementation.</p>
          </note>
        </item>
        <tag><marker id="MYtp"/><c><![CDATA[+MYtp <size>]]></c></tag>
        <item>
          <p>Top pad size (in kilobytes). This is the amount of extra
            memory that is allocated by <c>malloc</c> when
            <c>sbrk</c> is called to get more memory from the operating
            system. Defaults to <c>0</c>.</p>
          <note>
            <p>This flag has effect only when the emulator is linked with
              the GNU C library, and uses its <c>malloc</c> implementation.</p>
          </note>
        </item>
      </taglist>
    </section>

    <section>
      <title>Flags for Configuration of Allocators Based on alloc_util</title>
      <p>If <c>u</c> is used as subsystem identifier (that is,
        <c><![CDATA[<S> = u]]></c>), all allocators based on
        <c>alloc_util</c> are effected. If <c>B</c>, <c>D</c>, <c>E</c>,
        <c>F</c>, <c>H</c>, <c>I</c>, <c>L</c>, <c>R</c>, <c>S</c>, <c>T</c>,
	<c>X</c> is used as subsystem identifier, only the specific allocator
	identifier is effected.</p>

      <taglist>
        <tag><marker id="M_acul"/><c><![CDATA[+M<S>acul <utilization>|de]]></c>
        </tag>
        <item>
          <p>Abandon carrier utilization limit. A valid
            <c><![CDATA[<utilization>]]></c> is an integer in the range
            <c>[0, 100]</c> representing utilization in percent. When a
            utilization value &gt; 0 is used, allocator instances
            are allowed to abandon multiblock carriers. If <c>de</c> (default
            enabled) is passed instead of a <c><![CDATA[<utilization>]]></c>,
            a recommended non-zero utilization value is used. The value
            chosen depends on the allocator type and can be changed between
            ERTS versions. Defaults to <c>de</c>, but this
            can be changed in the future.</p>
          <p>Carriers are abandoned when
            memory utilization in the allocator instance falls below the
            utilization value used. Once a carrier is abandoned, no new
            allocations are made in it. When an allocator instance gets an
            increased multiblock carrier need, it first tries to fetch an
            abandoned carrier from another allocator instance. If no abandoned
            carrier can be fetched, it creates a new empty carrier. When an
            abandoned carrier has been fetched, it will function as an ordinary
            carrier. This feature has special requirements on the
            <seecref marker="#M_as">allocation strategy</seecref> used. Only
            the strategies <c>aoff</c>, <c>aoffcbf</c>, <c>aoffcaobf</c>,
	    <c>ageffcaoff</c>m, <c>ageffcbf</c> and <c>ageffcaobf</c>
            support abandoned carriers.</p>
          <p>This feature also requires
            <seecref marker="#M_t">multiple thread specific instances</seecref>
            to be enabled. When enabling this feature, multiple thread-specific
            instances are enabled if not already enabled, and the
            <c>aoffcbf</c> strategy is enabled if the current strategy does not
            support abandoned carriers. This feature can be enabled on all
            allocators based on the <c>alloc_util</c> framework, except
            <c>temp_alloc</c> (which would be pointless).</p>
        </item>

	<tag><marker id="M_acfml"/><c><![CDATA[+M<S>acfml <bytes>]]></c>
        </tag>
        <item>
          <p>Abandon carrier free block min limit. A valid <c><![CDATA[<bytes>]]></c>
            is a positive integer representing a block size limit. The largest
	    free block in a carrier must be at least <c>bytes</c> large, for the
	    carrier to be abandoned. The default is zero but can be changed
	    in the future.</p>
	  <p>See also <seecref marker="#M_acul"><c>acul</c></seecref>.</p>
        </item>

        <tag><marker id="M_acnl"/><c><![CDATA[+M<S>acnl <amount>]]></c>
        </tag>
        <item>
          <p>Abandon carrier number limit. A valid <c><![CDATA[<amount>]]></c>
            is a positive integer representing max number of abandoned carriers per
	    allocator instance. Defaults to 1000 which will practically disable
	    the limit, but this can be changed in the future.</p>
	  <p>See also <seecref marker="#M_acul"><c>acul</c></seecref>.</p>
        </item>

        <tag><marker id="M_as"/>
          <c><![CDATA[+M<S>as bf|aobf|aoff|aoffcbf|aoffcaobf|ageffcaoff|ageffcbf|ageffcaobf|gf|af]]></c></tag>
        <item>
          <p>Allocation strategy. The following strategies are valid:</p>
          <list type="bulleted">
            <item><c>bf</c> (best fit)</item>
            <item><c>aobf</c> (address order best fit)</item>
            <item><c>aoff</c> (address order first fit)</item>
            <item><c>aoffcbf</c> (address order first fit carrier best fit)
            </item>
            <item><c>aoffcaobf</c> (address order first fit carrier address
              order best fit)</item>
            <item><c>ageffcaoff</c> (age order first fit carrier address order first fit)</item>
            <item><c>ageffcbf</c> (age order first fit carrier best fit)
            </item>
            <item><c>ageffcaobf</c> (age order first fit carrier address
              order best fit)</item>
            <item><c>gf</c> (good fit)</item>
            <item><c>af</c> (a fit)</item>
          </list>
          <p>See the description of allocation strategies in section
             <seecref marker="#strategy">The alloc_util Framework</seecref>.</p>
        </item>
        <tag><marker id="M_asbcst"/><c><![CDATA[+M<S>asbcst <size>]]></c></tag>
        <item>
          <p>Absolute singleblock carrier shrink threshold (in
            kilobytes). When a block located in an
            <c>mseg_alloc</c> singleblock carrier is shrunk, the carrier
            is left unchanged if the amount of unused memory is less
            than this threshold, otherwise the carrier is shrunk.
            See also <seecref marker="#M_rsbcst"><c>rsbcst</c></seecref>.</p>
        </item>
        <tag><marker id="M_atags"/><c><![CDATA[+M<S>atags true|false]]></c></tag>
        <item>
          <p>Adds a small tag to each allocated block that contains basic
            information about what it is and who allocated it. Use the
            <seeerl marker="tools:instrument"><c>instrument</c></seeerl>
            module to inspect this information.</p>

          <p>The runtime overhead is one word per allocation when enabled. This
            may change at any time in the future.</p>

          <p>The default is <c>true</c> for <c>binary_alloc</c> and
            <c>driver_alloc</c>, and <c>false</c> for the other allocator
            types.</p>
        </item>
        <tag><marker id="M_cp"/><c><![CDATA[+M<S>cg B|D|E|F|H||L|R|S|@|:]]></c></tag>
        <item>
          <p>
	    Set carrier pool to use for the allocator. Memory carriers will
	    only migrate between allocator instances that use the same carrier
	    pool. The following carrier pool names exist:
	  </p>
	  <taglist>
	    <tag><c>B</c></tag>
	    <item>Carrier pool associated with <c>binary_alloc</c>.</item>
	    <tag><c>D</c></tag>
	    <item>Carrier pool associated with <c>std_alloc</c>.</item>
	    <tag><c>E</c></tag>
	    <item>Carrier pool associated with <c>ets_alloc</c>.</item>
	    <tag><c>F</c></tag>
	    <item>Carrier pool associated with <c>fix_alloc</c>.</item>
	    <tag><c>H</c></tag>
	    <item>Carrier pool associated with <c>eheap_alloc</c>.</item>
	    <tag><c>L</c></tag>
	    <item>Carrier pool associated with <c>ll_alloc</c>.</item>
	    <tag><c>R</c></tag>
	    <item>Carrier pool associated with <c>driver_alloc</c>.</item>
	    <tag><c>S</c></tag>
	    <item>Carrier pool associated with <c>sl_alloc</c>.</item>
	    <tag><c>@</c></tag>
	    <item>Carrier pool associated with the system as a whole.</item>
	  </taglist>
	  <p>
	    Besides passing carrier pool name as value to the parameter,
	    you can also pass <c>:</c>. By passing <c>:</c> instead of carrier
	    pool name, the allocator will use the carrier pool associated
	    with itself. By passing the command line argument "<c>+Mucg :</c>",
	    all allocators that have an associated carrier pool will use the
	    carrier pool associated with themselves.
	  </p>
	  <p>
	    The association between carrier pool and allocator is very loose.
	    The associations are more or less only there to get names for the
	    amount of carrier pools needed and names of carrier pools that
	    can be easily identified by the <c>:</c> value.
	  </p>
	  <p>
	    This flag is only valid for allocators that have an associated
	    carrier pool. Besides that, there are no restrictions on carrier
	    pools to use for an allocator.
	  </p>
	  <p>
	    Currently each allocator with an associated carrier pool
	    defaults to using its own associated carrier pool.
	  </p>
        </item>
        <tag><marker id="M_e"/><c><![CDATA[+M<S>e true|false]]></c></tag>
        <item>
          <p>Enables allocator <c><![CDATA[<S>]]></c>.</p>
        </item>
        <tag><marker id="M_lmbcs"/><c><![CDATA[+M<S>lmbcs <size>]]></c></tag>
        <item>
          <p>Largest (<c>mseg_alloc</c>) multiblock carrier size (in kilobytes).
            See the description on how sizes for <c>mseg_alloc</c> multiblock
            carriers are decided in section
            <seecref marker="#mseg_mbc_sizes">
            The alloc_util Framework</seecref>. On
            32-bit Unix style OS this limit cannot be set &gt; 64 MB.</p>
        </item>
        <tag><marker id="M_mbcgs"/><c><![CDATA[+M<S>mbcgs <ratio>]]></c></tag>
        <item>
          <p>(<c>mseg_alloc</c>) multiblock carrier growth stages.
            See the description on how sizes for <c>mseg_alloc</c> multiblock
            carriers are decided in section
            <seecref marker="#mseg_mbc_sizes">
            The alloc_util Framework</seecref>.</p>
        </item>
        <tag><marker id="M_mbsd"/><c><![CDATA[+M<S>mbsd <depth>]]></c></tag>
        <item>
          <p>Maximum block search depth. This flag has effect only if the
            good fit strategy is selected for allocator
            <c><![CDATA[<S>]]></c>. When the good fit strategy is used, free
            blocks are placed in segregated free-lists. Each free-list
            contains blocks of sizes in a specific range. The maxiumum block
            search depth sets a limit on the maximum number of blocks to
            inspect in a free-list during a search for suitable block
            satisfying the request.</p>
        </item>
        <tag><marker id="M_mmbcs"/><c><![CDATA[+M<S>mmbcs <size>]]></c></tag>
        <item>
          <p>Main multiblock carrier size. Sets the size of the main
            multiblock carrier for allocator <c><![CDATA[<S>]]></c>. The main
            multiblock carrier is allocated through <c><![CDATA[sys_alloc]]></c>
            and is never deallocated.</p>
        </item>
        <tag><marker id="M_mmmbc"/><c><![CDATA[+M<S>mmmbc <amount>]]></c></tag>
        <item>
          <p>Maximum <c>mseg_alloc</c> multiblock carriers. Maximum number of
            multiblock carriers allocated through <c>mseg_alloc</c> by
            allocator <c><![CDATA[<S>]]></c>. When this limit is reached,
            new multiblock carriers are allocated through
            <c>sys_alloc</c>.</p>
        </item>
        <tag><marker id="M_mmsbc"/><c><![CDATA[+M<S>mmsbc <amount>]]></c></tag>
        <item>
          <p>Maximum <c>mseg_alloc</c> singleblock carriers. Maximum number of
            singleblock carriers allocated through <c>mseg_alloc</c> by
            allocator <c><![CDATA[<S>]]></c>. When this limit is reached,
            new singleblock carriers are allocated through
            <c>sys_alloc</c>.</p>
        </item>
        <tag><marker id="M_ramv"/><c><![CDATA[+M<S>ramv <bool>]]></c></tag>
        <item>
          <p>Realloc always moves. When enabled, reallocate operations are
            more or less translated into an allocate, copy, free sequence.
            This often reduces memory fragmentation, but costs performance.</p>
        </item>
        <tag><marker id="M_rmbcmt"/><c><![CDATA[+M<S>rmbcmt <ratio>]]></c></tag>
        <item>
          <p>Relative multiblock carrier move threshold (in percent). When
            a block located in a multiblock carrier is shrunk,
            the block is moved if the ratio of the size of the returned
            memory compared to the previous size is more than this threshold,
            otherwise the block is shrunk at the current location.</p>
        </item>
        <tag><marker id="M_rsbcmt"/><c><![CDATA[+M<S>rsbcmt <ratio>]]></c></tag>
        <item>
          <p>Relative singleblock carrier move threshold (in percent). When
            a block located in a singleblock carrier is shrunk to
            a size smaller than the value of parameter
            <seecref marker="#M_sbct"><c>sbct</c></seecref>,
            the block is left unchanged in the singleblock carrier if
            the ratio of unused memory is less than this threshold,
            otherwise it is moved into a multiblock carrier.</p>
        </item>
        <tag><marker id="M_rsbcst"/><c><![CDATA[+M<S>rsbcst <ratio>]]></c></tag>
        <item>
          <p>Relative singleblock carrier shrink threshold (in
            percent). When a block located in an <c>mseg_alloc</c>
            singleblock carrier is shrunk, the carrier is left
            unchanged if the ratio of unused memory is less than this
            threshold, otherwise the carrier is shrunk.
            See also <seecref marker="#M_asbcst"><c>asbcst</c></seecref>.</p>
        </item>
        <tag><marker id="M_sbct"/><c><![CDATA[+M<S>sbct <size>]]></c></tag>
        <item>
          <p>Singleblock carrier threshold (in kilobytes). Blocks larger than this
            threshold are placed in singleblock carriers. Blocks
            smaller than this threshold are placed in multiblock
            carriers. On 32-bit Unix style OS this threshold cannot be set
            &gt; 8 MB.</p>
        </item>
        <tag><marker id="M_smbcs"/><c><![CDATA[+M<S>smbcs <size>]]></c></tag>
        <item>
          <p>Smallest (<c>mseg_alloc</c>) multiblock carrier size (in
            kilobytes). See the description on how sizes for <c>mseg_alloc</c>
            multiblock carriers are decided in section
            <seecref marker="#mseg_mbc_sizes">
            The alloc_util Framework</seecref>.</p>
        </item>
        <tag><marker id="M_t"/><c><![CDATA[+M<S>t true|false]]></c></tag>
        <item>
          <p>Multiple, thread-specific instances of the allocator.
            This option has only effect on the runtime system
            with SMP support. Default behavior on the runtime system with
            SMP support is <c>NoSchedulers+1</c> instances. Each scheduler
            uses a lock-free instance of its own and other threads use
            a common instance.</p>
          <p>Before ERTS 5.9 it was possible to configure
            a smaller number of thread-specific instances than schedulers.
            This is, however, not possible anymore.</p>
        </item>
      </taglist>
    </section>

    <section>
      <title>Flags for Configuration of alloc_util</title>
      <p>All allocators based on <c>alloc_util</c> are effected.</p>

      <taglist>
        <tag><marker id="Muycs"/><c><![CDATA[+Muycs <size>]]></c></tag>
        <item>
          <p><c>sys_alloc</c> carrier size. Carriers allocated through
            <c>sys_alloc</c> are allocated in sizes that are
            multiples of the <c>sys_alloc</c> carrier size. This is not
            true for main multiblock carriers and carriers allocated
            during a memory shortage, though.</p>
        </item>
        <tag><marker id="Mummc"/><c><![CDATA[+Mummc <amount>]]></c></tag>
        <item>
          <p>Maximum <c>mseg_alloc</c> carriers. Maximum number of carriers
            placed in separate memory segments. When this limit is
            reached, new carriers are placed in memory retrieved from
            <c>sys_alloc</c>.</p>
        </item>
        <tag><marker id="Musac"/><c><![CDATA[+Musac <bool>]]></c></tag>
        <item>
          <p>Allow <c>sys_alloc</c> carriers. Defaults to <c>true</c>.
            If set to <c>false</c>, <c>sys_alloc</c> carriers are never
            created by allocators using the <c>alloc_util</c> framework.</p>
        </item>
      </taglist>
    </section>

    <section>
      <title>Special Flag for literal_alloc</title>
      <taglist>
        <tag><marker id="MIscs"/><c><![CDATA[+MIscs <size in MB>]]></c></tag>
        <item>
          <p><c>literal_alloc</c> super carrier size (in MB). The amount of
            <em>virtual</em> address space reserved for literal terms in
            Erlang code on 64-bit architectures. Defaults to <c>1024</c>
            (that is, 1 GB), which is usually sufficient.
            The flag is ignored on 32-bit architectures.</p>
        </item>
      </taglist>
    </section>

    <section>
      <title>Instrumentation Flags</title>
      <taglist>
        <tag><c>+M&lt;S&gt;atags</c></tag>
        <item>
          <p>Adds a small tag to each allocated block that contains basic
            information about what it is and who allocated it. See
            <seecref marker="#M_atags"><c>+M&lt;S&gt;atags</c></seecref> for a
            more complete description.</p>
        </item>
        <tag><marker id="Mit"/><c>+Mit X</c></tag>
        <item>
          <p>Reserved for future use. Do <em>not</em> use this flag.</p>
        </item>
      </taglist>

      <note>
        <p>When instrumentation of the emulator is enabled, the emulator
          uses more memory and runs slower.</p>
      </note>
    </section>

    <section>
      <title>Other Flags</title>
      <taglist>
        <tag><marker id="Mea"/><c>+Mea min|max|r9c|r10b|r11b|config</c></tag>
        <item>
          <p>Options:</p>
          <taglist>
            <tag><c>min</c></tag>
            <item>
              <p>Disables all allocators that can be disabled.</p>
            </item>
            <tag><c>max</c></tag>
            <item>
              <p>Enables all allocators (default).</p>
            </item>
            <tag><c>r9c|r10b|r11b</c></tag>
            <item>
              <p>Configures all allocators as they were configured in respective
                Erlang/OTP release. These will eventually be removed.</p>
            </item>
            <tag><c>config</c></tag>
            <item>
              <p>Disables features that cannot be enabled while creating an
                allocator configuration with
                <seeerl marker="runtime_tools:erts_alloc_config">
                <c>erts_alloc_config(3)</c></seeerl>.</p>
              <note>
                <p>This option is to be used only while running
                  <c>erts_alloc_config(3)</c>, <em>not</em> when
                  using the created configuration.</p>
              </note>
            </item>
          </taglist>
        </item>
        <tag><marker id="Mlpm"/><c>+Mlpm all|no</c></tag>
        <item>
          <p>Lock physical memory. Defaults to <c>no</c>, that is,
            no physical memory is locked. If set to <c>all</c>, all
            memory mappings made by the runtime system are locked into
            physical memory. If set to <c>all</c>, the runtime system fails to
            start if this feature is not supported, the user has not got enough
            privileges, or the user is not allowed to lock enough physical
            memory. The runtime system also fails with an out of memory
            condition if the user limit on the amount of locked memory is
            reached.</p>
        </item>
      </taglist>
    </section>
  </section>

  <section>
    <title>Notes</title>
    <p>Only some default values have been presented here. For information
      about the currently used settings and the current status of the
      allocators, see
      <seeerl marker="erts:erlang#system_info_allocator">
      <c>erlang:system_info(allocator)</c></seeerl> and
      <seeerl marker="erts:erlang#system_info_allocator_tuple">
      <c>erlang:system_info({allocator, Alloc})</c></seeerl>.</p>

    <note>
      <p>Most of these flags are highly implementation-dependent and
        can be changed or removed without prior notice.</p>
      <p><c>erts_alloc</c> is not obliged to strictly use the settings that
        have been passed to it (it can even ignore them).</p>
    </note>

    <p>The <seeerl marker="runtime_tools:erts_alloc_config">
      <c>erts_alloc_config(3)</c></seeerl>
      tool can be used to aid creation of an
      <c>erts_alloc</c> configuration that is suitable for a limited
      number of runtime scenarios.</p>
  </section>

  <section>
    <title>See Also</title>
    <p><seecom marker="erl"><c>erl(1)</c></seecom>,
      <seeerl marker="erlang"><c>erlang(3)</c></seeerl>,
      <seeerl marker="runtime_tools:erts_alloc_config">
      <c>erts_alloc_config(3)</c></seeerl>,
      <seeerl marker="tools:instrument">
      <c>instrument(3)</c></seeerl></p>
  </section>
</cref>